Phonetic Dimensions Research Articles

Formant and Voice Source Characteristics of Vowels in Chinese National Singing and Bel Canto. A Pilot Study

There have been numerous reports on the acoustic characteristics of singers' vowel articulation and phonation, and these studies cover many phonetic dimensions, such as fundamental frequency (F0), intensity, formant frequency, and voice quality. Taking the three representative vowels (/a/, /i/, /u/) in Chinese National Singing and Bel Canto as the research object, the present study investigates the differences and associations in vowel articulation and phonation between Chinese National Singing and Bel Canto using acoustic measures, for example, F0, formant frequency, long-term average spectrum (LTAS). The relationship between F0 and formant indicates that F1 is proportional to F0, in which the female has a significant variation in vowel /a/. Compared with the male, the formant structure of the female singing voice differs significantly from that of the speech voice. Regarding the relationship between intensity and formant, LTAS shows that the Chinese National Singing tenor and Bel Canto baritone have the singer's formant cluster when singing vowels, while the two sopranos do not. The systematic changes of formant frequencies with voice source are observed. (i) F1 of the female vowel /a/ has undergone a significant tuning change in the register transition, reflecting the characteristics of singing genres. (ii) Female singers utilize the intrinsic pitch of vowels when adopting the register transition strategy. This finding can be assumed to facilitate understanding the theory of intrinsic vowel pitch and revise Sundberg's hypothesis that F1 rises with F0. A non-linear relationship exists between F1 and F0, which adds to the non-linear interaction of the formant and vocal source. (iii) The singer's formant is affected by voice classification, gender, and singing genres.

Revealing perceptual structure through input variation: cross-accent categorization of vowels in five accents of English

This paper characterizes the perceptual structure of vowel systems in five regional accents of English, from Australia (A), New Zealand (Z), London (L), Yorkshire (Y), and Newcastle upon Tyne (N), on the basis of “whole system” vowel categorization experiments. We established patterns of within-accent vowel confusions, and then explored cross-accent perception, assessing how listeners from one accent background categorize vowels from another. Our experimental task required mapping continuous phonetic dimensions to perceptual categories in the absence of phonotactic and lexical cues to vowel identity and socio-indexical information about the talker. Our results show that, without these sources of information, there is uncertainty in vowel categorization, even for native accent vowels, and that this degree of uncertainty increases for unfamiliar accents. The patterns of cross-accent perception largely reflect the accent-specific perceptual structure of the listener, as opposed to adaptations to the stimulus accents. This finding contrasts with the type of active talker adaptation found with tasks offering lexical information about vowel identity and indexical information about the talker.

Native and Non-Native Accents of English: Different Perceptions of English as a Global Language in a Phonetic Perspective. An Editorial to RiL Special Issue Vol. 20.1 and 20.2.

The text reviews selected current issues related to native and non-native accents of English, including the status of English as a global language, the value of varieties of English and the question of a possible supremacy of one accent over other varieties. The text links theoretical and applied issues addressing the question of which variety should be used in EFL education. Focusing on the phonetic dimension it weights the conflicting reasons given in classic works, such as Kachru’s and McArthur’s to eventually point to novel research programmes as exemplified in other texts included in the special issues of Research in Language, i.e. vol. 20, issues 1 and 2, most of which were inspired by Accents 2021, an annual conference organised by the Department of English Language and Applied Linguistics, Institute of English Studies, University of Lodz, Poland.

Learning from a single cue: Is phonetic learning dimension-based?

Phonetic cue-weighting, the process of altering the weights of certain dimensions (e.g., F0) in the speech signal, is a fundamental process in speech perception. Cue-reweighting is the process of adaptation required for understanding new accents and learning second language speech contrasts; however, little is understood about the underlying mechanisms. Harmon et al. (2019) examined three candidate mechanisms (distributional, supervised, and reinforcement learning) showing evidence for reinforcement learning. The current study investigates Harmon et al.’s (2019) assumed phonetic dimensions by asking how a single cue in a phonetic dimension (e.g., a single voice onset time (VOT) value) of a phonological contrast ([b]/[p]) generalizes to other values of the phonetic dimension. Said simpler, is phonetic learning dimension-based? Native English listeners (N = 270) participated in an online perceptual training experiment in which participants were asked to identify word contrasts like pear and bear. Results suggest that learning to downweigh a cue (e.g., VOT = 5 ms) for [b]/[p] generalizes across new VOT values (e.g., VOT = 15 ms). However, the generalization did not extend to the most distant value (e.g., VOT = 35 ms). That is, cue-reweighting can affect a single phonetic category but does not extend to the entire phonetic dimension across category boundaries.

Uniformity in phonetic realization: Evidence from sibilant place of articulation in American English

UNIFORMITY IN PHONETIC REALIZATION: EVIDENCE FROM SIBILANT PLACE OF ARTICULATION IN AMERICAN ENGLISH ELEANOR CHODROFF COLIN WILSON University of York Johns Hopkins University Phonetic realization is highly variable and highly structured within and across talkers. We examine three constraints that could structure the phonetic space of related speech sounds: target, contrast, and pattern uniformity. Target uniformity requires a uniform mapping from distinctive features to their corresponding phonetic targets within a talker, contrast uniformity requires a consistent difference in the phonetic targets that realize featural contrasts across talkers, and pattern uniformity requires a uniform template of phonetic targets across talkers. Focusing on American English sibilant fricatives, we measure and compare each constraint’s influence on the phonetic targets corresponding to place of articulation. We find that target uniformity is the strongest constraint : each talker realizes a given distinctive feature value in highly similar ways across related sounds. Together with similar findings for other sound classes, this result reveals fine-grained systematicity in the mapping from phonology to phonetics and has implications for theories of speech production and speech perception.* Keywords: phonetic realization, sibilant fricatives, uniformity, talker variability, Bayesian models 1. Introduction. No one-to-one mapping exists between linguistic units and their phonetic instantiations (Liberman et al. 1967, Massaro 1975, Pisoni & Sawusch 1975). This lack of invariance is a fundamental issue for both the perception and production of language. From the perspective of perception, how do perceivers adapt to extensive variation in the physical signal (whether spoken or signed)? From the perspective of production, how do producers know the limits of acceptable variation for their particular language variety, or even just for intelligibility? It is well established that variation in phonetic realization is extensive yet structured in many ways (Labov 1972, Miller 1994, Foulkes et al. 2001, Kleinschmidt & Jaeger 2015, Guy & Hinskens 2016, Sonderegger et al. 2020). In the present article, we explore potential constraints on the mapping from phonological representations, such as segments and their distinctive features, to targets of phonetic realization. We begin by considering a subinventory of two or more related sounds (e.g. [i] and [u], or [s] and [z]) and their corresponding phonetic targets (i.e. perceptuomotor representations ). A given talker could structure the phonetic realization of these sounds by copying a pattern or template of targets that exists in the speech community and adapting it to their anatomy. This scenario allows for talker variation—one speaker’s realization of the template may be overall higher or lower on a given phonetic dimension—but otherwise it can be construed as ‘maximal phonetic structure’. Provided that the hypothetical template can be adapted to each speaker’s anatomy, this system would be fully general across the speaker population. Moreover, clear motivation for such a system exists in speech perception : if each talker has the same template of phonetic targets, perceptual adaptation would involve a simple translation of the pattern for each new talker. This is in fact assumed by many approaches to talker normalization and adaptation, especially for vowel systems (e.g. Lobanov 1971, Nearey 1978, Nearey & Assmann 2007). 1 * The authors would like to thank Shravan Vasishth for hosting the 2020 Potsdam Summer School on Statistical Methods in Linguistics and Psychology, Lisa Davidson for sharing the laboratory data, and Ryan Cotterell , Matthew Faytak, Josef Fruehwald, and Jane Stuart-Smith for helpful discussion. All data and analyses are available at https://osf.io/bysfa/. Printed with the permission of Eleanor Chodroff & Colin Wilson. © 2022. In opposition to maximal phonetic structure in the speech inventory, we can consider ‘maximal phonetic bricolage’. Bricolage reflects the constellation of linguistic variables that a talker can exploit for expressing social identity (Eckert 2008, Zimman 2017); taken to the extreme, it would allow talkers to pick and choose phonetic targets independently for each sound. In this scenario, the phonetic space may be structured by overarching social variables, but it would be entirely unstructured within the subinventory and across speakers. For example, the relationship among the phonetic targets of sibilant fricatives like [s], [z], [ʃ], and [ʒ] could be different for each speaker, depending on how each target is chosen to express some aspect of social identity. Existing evidence points to an intermediate scenario between these two endpoints...

Contrast implementation affects phonetic variability: A case study of Hindi and English stops

There is a large body of work in phonetics and phonology demonstrating sources and structure of acoustic variability, showing that variability in speech production is not random. This paper examines the question of how variability itself varies across languages and speakers, arguing that differences in extent of variability are also systematic. A classic hypothesis from Dispersion Theory (Lindblom, 1986) posits a relationship between extent of variability and phoneme inventory size, but this has been shown to be inadequate for predicting differences in phonetic variability. I propose an alternative hypothesis, Contrast-Dependent Variation, which considers cue weight of individual phonetic dimensions rather than size of phonemic inventories. This is applied to a case study of Hindi and American English stops and correctly predicts more variability in English stop closure voicing relative to Hindi, but similar amounts of lag time variability in both languages. In addition to these group-level between- language differences, the results demonstrate how patterns of individual speaker differences are language-specific and conditioned by differences in phonological contrast implementation.

Phonetic adaptation in interlocutors with mismatched language backgrounds: A case for a phonetic synergy account

Phonetic convergence is the phenomenon that the speech of interlocutors becomes more similar during an interaction. However, in interactions involving non-native speakers, factors such as communicative needs and mismatches in interlocutors’ phonetic repertoire may inhibit phonetic convergence. In this study, we investigated how mismatches at the phonetic level affect speech adjustments in mixed versus shared language pairs. Mandarin L1 speakers speaking English as their second language played a collaborative word matching game either with another Mandarin L1–English L2 speaker (MM pair) or with a monolingual English speaker (ME pairs). Critical stimuli consisted of single-syllable words that differed on medial vowel and/or voicing of the final consonant. These phonetic dimensions were chosen because Mandarin L1–English L2 speakers differ from native English speakers in how they instantiate these distinctions. We evaluated how the different kinds of pairs completed the interactive communicative task.Our results confirmed that L2 speakers differed from L1 speakers in how they created necessary phonetic distinctions. However, these differences depended on whether they were in an MM pair or an ME pair. L2 speakers in MM pairs relied on vowel duration to indicate vowel identity more than did those in the ME pairs. Interestingly, examination of individual speakers’ data revealed different successful solutions across pairs. To explain these findings, we propose the use of phonetic synergy the idea that interlocutors’ adaptations to ensure communicative success are co-constraining and diverse. Broadly, our findings suggest that for a complete understanding of phonetic convergence, precisely manipulating communicative difficulty along specific phonetic dimensions may be crucial to tell apart social, simulative, and communicative adjustments in interaction.

Phones in larger inventories are not (necessarily) less variable

It is often assumed that there should be relatively less within-category phonetic variability in production in languages which have larger phonemic inventories (Lindblom, 1986 on Dispersion Theory). Although this hypothesis is intuitive, existing evidence in the literature is mixed and suggests the need to modulate the hypothesis according to phonological context (e.g., Renwick, 2012) and individual cues (Recasens and Espinosa, 2006). In this paper, we examine this prediction using data from sibilant fricatives in French, Polish, and Mandarin. A direct implementation of Lindblom (1986) predicts more variation in French (2 sibilants) relative to Polish and Mandarin (3 sibilants). We obtained data from these languages in a speech production experiment, analyzing within-category variability in sibilant center of gravity and onset of the second formant of the following vowel. We observed no language-specific differences in extent of within-category variability on either phonetic dimension. We did, however, observe systematic individual differences in extent of within-category variability among speakers of the same language, which were correlated with individual differences in production cue weight. We interpret these results as evidence that extent of variation is related to phonological contrast, but contrast must be quantified in terms of cue weight, rather than phoneme inventory size.

Prosody leaks into the memories of words

The average predictability (aka informativity) of a word in context has been shown to condition word duration (Seyfarth, 2014). All else being equal, words that tend to occur in more predictable environments are shorter than words that tend to occur in less predictable environments. One account of the informativity effect on duration is that the acoustic details of probabilistic reduction are stored as part of a word's mental representation. Other research has argued that predictability effects are tied to prosodic structure in integral ways. With the aim of assessing a potential prosodic basis for informativity effects in speech production, this study extends past work in two directions; it investigated informativity effects in another large language, Mandarin Chinese, and broadened the study beyond word duration to additional acoustic dimensions, pitch and intensity, known to index prosodic prominence. The acoustic information of content words was extracted from a large telephone conversation speech corpus with over 400,000 tokens and 6000 word types spoken by 1655 individuals and analyzed for the effect of informativity using frequency statistics estimated from a 431 million word subtitle corpus. Results indicated that words with low informativity have shorter durations, replicating the effect found in English. In addition, informativity had significant effects on maximum pitch and intensity, two phonetic dimensions related to prosodic prominence. Extending this interpretation, these results suggest that predictability is closely linked to prosodic prominence, and that the lexical representation of a word includes phonetic details associated with its average prosodic prominence in discourse. In other words, the lexicon absorbs prosodic influences on speech production.

Redefining Sociophonetic Competence: Mapping COG Differences in Phrase-Final Fricative Epithesis in L1 and L2 Speakers of French

This article presents a study of measures of center of gravity (COG) in phrase-final fricative epithesis (PFFE) produced by L1 and L2 speakers of Continental French (CF). Participants completed a reading task targeting 98 tokens of /i,y,u/ in phrase-final position. COG measures were taken at the 25%, 50% and 75% marks, normalized and submitted to a mixed linear regression. Results revealed that L2 speakers showed higher COG values than L1 speakers in low PFFE-to-vowel ratios at the 25%, 50%, and 75% marks. COG measures were then categorized into six profile types on the basis of their frequencies at each timepoint: flat–low, flat–high, rising, falling, rising–falling, and falling–rising. Counts of COG profile were then submitted to multinomial logistic regression. Results revealed that although L1 speakers produced predominantly flat–low profile types at lower percent devoicings, L2 speakers preferred multiple strategies involving higher levels of articulatory energy (rising, falling, rise–fall). These results suggest that while L1 speakers realize PFFE differently with respect to phonological context, L2 speakers rely on its most common allophone, strong frication, in most contexts. As such, the findings of this study argue for an additional phonetic dimension in the construct of L2 sociophonetic competence.

From sounds to words: The relation between phonological and lexical processing of tone in L2 Mandarin

Successful listening in a second language (L2) involves learning to identify the relevant acoustic–phonetic dimensions that differentiate between words in the L2, and then use these cues to access lexical representations during real-time comprehension. This is a particularly challenging goal to achieve when the relevant acoustic–phonetic dimensions in the L2 differ from those in the L1, as is the case for the L2 acquisition of Mandarin, a tonal language, by speakers of non-tonal languages like English. Previous work shows tone in L2 is perceived less categorically (Shen and Froud, 2019) and weighted less in word recognition (Pelzl et al., 2019) than in L1. However, little is known about the link between categorical perception of tone and use of tone in real time L2 word recognition at the level of the individual learner. This study presents evidence from 30 native and 29 L1-English speakers of Mandarin who completed a real-time spoken word recognition and a tone identification task. Results show that L2 learners differed from native speakers in both the extent to which they perceived tone categorically as well as in their ability to use tonal cues to distinguish between words in real-time comprehension. Critically, learners who reliably distinguished between words differing by tone alone in the word recognition task also showed more categorical perception of tone on the identification task. Moreover, within this group, performance on the two tasks was strongly correlated. This provides the first direct evidence showing that the ability to perceive tone categorically is related to the weighting of tonal cues during spoken word recognition, thus contributing to a better understanding of the link between phonemic and lexical processing, which has been argued to be a key component in the L2 acquisition of tone (Wong and Perrachione, 2007).

Correlation Between Car Size, Weight, Power, and Vowel Quality in Model Names

Abstract This paper focuses on the practical application of the theory of sound symbolism in brand name development and examines which of the two phonetic dimensions of vowel articulation, the vertical articulatory scale or the horizontal one, is utilised to a higher degree in communicating the size of a vehicle to customers. The methodology used in previous studies on size-sound symbolism did not make it possible to separate the two aspects of vowel articulation. In the present paper, these dimensions were categorised by the use of quantitative methods. Each Received Pronunciation vowel was assigned a numerical value separately on both scales. Then, the correlations between the values obtained for horizontal and vertical articulation of the vowels present in the names of cars sold in Great Britain and the physical attributes of the respective vehicles such as size, weight, and power were calculated. The final results reveal that it is only the vertical scale of vowel articulation which is utilised to signal the physical characteristics of the vehicles examined in this project. Although these findings refer directly to British English, they may also have more universal implications for the theory of magnitude sound symbolism.

Acoustic\u2013phonetic and auditory mechanisms of adaptation in the perception of sibilant fricatives

Listeners are highly proficient at adapting to contextual variation when perceiving speech. In the present study, we examined the effects of brief speech and nonspeech contexts on the perception of sibilant fricatives. We explored three theoretically motivated accounts of contextual adaptation, based on phonetic cue calibration, phonetic covariation, and auditory contrast. Under the cue calibration account, listeners adapt by estimating a talker-specific average for each phonetic cue or dimension; under the cue covariation account, listeners adapt by exploiting consistencies in how the realization of speech sounds varies across talkers; under the auditory contrast account, adaptation results from (partial) masking of spectral components that are shared by adjacent stimuli. The spectral center of gravity, a phonetic cue to fricative identity, was manipulated for several types of context sound: /z/-initial syllables, /v/-initial syllables, and white noise matched in long-term average spectrum (LTAS) to the /z/-initial stimuli. Listeners’ perception of the /s/–/ʃ/ contrast was significantly influenced by /z/-initial syllables and LTAS-matched white noise stimuli, but not by /v/-initial syllables. No significant difference in adaptation was observed between exposure to /z/-initial syllables and matched white noise stimuli, and speech did not have a considerable advantage over noise when the two were presented consecutively within a context. The pattern of findings is most consistent with the auditory contrast account of short-term perceptual adaptation. The cue covariation account makes accurate predictions for speech contexts, but not for nonspeech contexts or for the absence of a speech-versus-nonspeech difference.

Phonetic cue weighting in perception and production.

Speech sound contrasts differ along multiple phonetic dimensions. During speech perception, listeners must decide which cues are relevant, and determine the relative importance of each cue, while also integrating other, signal-external cues. The comparison of cue weighting in perception and production bears on a range of theoretical issues including the processes underlying sound change, the time course of learning, the nature of cues, and the perception-production interface. Research examining the relative alignment of cue weighting across the modalities, on both a community and individual level, has revealed both parallels and asymmetries between the modalities. The extraordinarily wide range of ways that have been used to conceptualize and quantify cue weights reflects the inherent theoretical, methodological, and analytical differences between the two modalities. More consideration of the choices of analytical metrics, explicit discussion of the theoretical assumptions that underlie them, and systematic investigations of different types of cues will lead to more generalizable findings that can be incorporated into computational implementable models of speech processing. This article is categorized under: Linguistics > Language in Mind and Brain Psychology > Language.

La marca: un signo de identificación visual y auditivo sinérgico

A partir del hecho de considerar a la marca como un signo de identificación visual yauditivo sinérgico, se observa que, frente a la presencia de la estructura gráfica, emerge indefectiblemente el nombre de ésta, es decir aquellos elementos “pronunciables”.Excediendo el campo de la comunicación visual y el diseño gráfico, se suscita entonces la necesidad de incorporar al análisis de la marca- signo, la dimensión fonética de la misma.En la marca verbal intervienen ‘letras’ o más específicamente, fonemas (modelos o tipos idealesde sonidos) con su particular connotación fonética. Es decir que las vocales, las consonantesy los grupos fonéticos presentan un determinado valor simbólico. La interpretación de los fonosímbolos permite trascender más allá de la imagen acústica, incluyendo el estudio de losfonemas expresados verbalmente en el nombre de la marca, y la capacidad de los sonidos deevocar o sugerir un concepto, una idea, una emoción o una acción.

Mutual predictability among speech sounds for talker adaptation and recognition

Processes of talker recognition and adaptation rely on a high degree of inter-talker phonetic variability and systematicity, respectively. While superficially in opposition, talker recognition in part depends on adaptation to the talker at hand. In this talk, we present evidence that talker variability is simultaneously extensive and structured within natural classes of speech sounds. In American English, talker mean peak frequencies for [s] span over 3000 Hz, but the variation in [s] is not independent of that in [z]: strong correlations of the talker mean peak frequency, among other phonetic dimensions, are observed between sibilant fricatives. Covariation among speech sounds indicates mutual predictability, such that evidence from one speech sound could be used to refine estimates or make predictions about a second. Listeners indeed demonstrate perceptual knowledge of covariation in generalized adaptation to novel talkers. After exposure to a talker with a relatively high- or low-peak frequency [z], listeners adjusted their [s]-[ʃ] boundary in accordance with the empirical covariation. As talker recognition entails estimation of a talker’s phonetic parameters, prior perceptual knowledge of covariation could be used to refine estimation of multiple speech sounds from minimal exposure, thus accelerating processes of talker adaptation and recognition.

Voice onset time and beyond: Exploring laryngeal contrast in 19 languages

In this special collection entitled Marking 50 Years of Research on Voice Onset Time and the Voicing Contrast in the World’s Languages, we have compiled eleven studies investigating the voicing contrast in 19 languages. The collection provides extensive data obtained from 270 speakers across those languages, examining VOT and other acoustic, aerodynamic and articulatory measures. The languages studied may be divided into four groups: ‘aspirating’ languages with a two-way contrast (English, three varieties of German); ‘true voicing’ languages with a two-way contrast (Russian, Turkish, Brazilian Portuguese, two Iranian languages Pashto and Wakhi); languages with a three-way contrast (Thai, Vietnamese, Khmer, Yerevan Armenia, three Indo-Aryan languages, Dawoodi, Punjabi and Shina, and Burushaki); and Indo-Aryan languages with a more than three-way contrast (Jangli and Urdu with a four-way contrast, and Sindhi and Siraiki with a five-way contrast). We discuss the cross-linguistic data, focusing on how much VOT alone tells us about the voicing contrast in these languages, and what other phonetic dimensions (such as consonant-induced F0 and voice quality) are needed for a complete understanding of laryngeal contrast in these languages. Implications for various issues emerge: universal phonetic feature systems, effects of language contact on linguistic levelling, and the relation between laryngeal contrast and supralaryngeal articulation. The cross-linguistic VOT data also lead us to discuss how the distribution of VOT as measured acoustically may allow us to infer the underlying articulation and how it might be approached in gestural phonologies. The discussion on these multiple issues sparks new questions to be resolved, and provide indications of where the field may be best directed in exploring laryngeal contrast in voicing in the world’s languages.

Production and Perception of F0 by Three-Year-Old Children

This study explores how young Korean children produce and perceive stop contrasts in the phonetic dimension of F0. The Korean spoken in Seoul is undergoing tonogenetic sound change in which there has been a loss of VOT differentiation in young adults’ speech, while the role of fundamental frequency (F0) in enhancing Korean stop contrasts has been required. To investigate the interrelation between children’s production and perception of F0, the experiments were conducted on three-year-old children. Phonetic analysis showed that lenis and aspirated stops have no significant VOT differences, whereas a significant relationship with F0 was found. The results suggest that the acquisition of F0 plays a crucial role in the formation of phonemic categories for lenis and aspirated stops and this process significantly affects articulatory distinction. (Kwangwoon University)

Stop, approximant, and timing slot: The changing faces of the velar stop in Iwaidja

Limited access to speakers and incomplete lexical knowledge are common challenges facing phonetic description of under-documented languages. We address these challenges by taking a multi-dimensional approach, seeking to constrain our phonetic description by covariation across acoustic and articulatory parameters. We demonstrate the approach through an analysis of velar consonsants in the Australian Aboriginal language Iwaidja. Existing accounts contrast a velar stop /k/ with a velar approximant /ɰ/ in word-medial position (Evans 2009). Converging evidence from ultrasound images of the tongue body and acoustic analysis of intensity data reveal that the posited opposition is not consistent across speakers (N = 4) and lexical items. Unsupervised categorization of the phonetic data indicates two phonetic categories, appropriately labelled as [a] and [ɰ], which do not map consistently to dictionary labels in existing descriptions. We conclude that speaker-specific allophonic variation is the result of an ongoing process of lenition of /k/ between sonorant segments which has not yet phonologized. More broadly, integrating phonetic dimensions revealed categories that were ill-defined on the basis of just acoustic or articulatory measures alone. Depth of analysis, characterized by phonetic multi-dimensionality, may support robust generalization where broad analysis (multiple speakers, large corpora) are impractical or impossible.Limited access to speakers and incomplete lexical knowledge are common challenges facing phonetic description of under-documented languages. We address these challenges by taking a multi-dimensional approach, seeking to constrain our phonetic description by covariation across acoustic and articulatory parameters. We demonstrate the approach through an analysis of velar consonsants in the Australian Aboriginal language Iwaidja. Existing accounts contrast a velar stop /k/ with a velar approximant /ɰ/ in word-medial position (Evans 2009). Converging evidence from ultrasound images of the tongue body and acoustic analysis of intensity data reveal that the posited opposition is not consistent across speakers (N = 4) and lexical items. Unsupervised categorization of the phonetic data indicates two phonetic categories, appropriately labelled as [a] and [ɰ], which do not map consistently to dictionary labels in existing descriptions. We conclude that speaker-specific allophonic variation is the result of an ongoi...

Assessing surface phonological specification through simulation and classification of phonetic trajectories

Many previous studies have argued that phonology may leave some phonetic dimensions unspecified in surface representations. We introduce computational tools for assessing this possibility though simulation and classification of phonetic trajectories. The empirical material used to demonstrate the approach comes from electromagnetic articulography recordings of high-vowel devoicing in Japanese. Using Discrete Cosine Transform, tongue-dorsum movement trajectories are decomposed into a small number of frequency components (cosines differing in frequency and amplitude) that correspond to linguistically meaningful signal modulations, i.e. articulatory gestures. Stochastic generators of competing phonological hypotheses operate in this frequency space. Distributions over frequency components are used to simulate (i) the vowel-present trajectories and (ii) the vowel-absent trajectories. A Bayesian classifier trained on simulations assigns posterior probabilities to unseen data. Results indicate that /u/ is optionally produced without a vowel-height specification in Tokyo Japanese and that the frequency of such targetlessness varies systematically across phonological environments.